diff --git a/docs/en/faq.md b/docs/en/faq.md index 9d33bb0..a475e7d 100644 --- a/docs/en/faq.md +++ b/docs/en/faq.md @@ -22,7 +22,16 @@ Yes, you can use traditional pip + venv approach. ### Q: Do I need to configure ComfyUI? -Not necessarily. You can use RunningHub cloud service without local deployment. +**Not necessarily** - it depends on your template choice: + +| Template Type | ComfyUI | Best For | Speed | +|--------------|---------|----------|-------| +| Text-only
(e.g., `simple.html`) | ❌ Not needed | Quotes, announcements, reading prompts | ⚡⚡⚡ Very fast | +| AI Images
(e.g., `default.html`) | ✅ Required | Rich visual content | ⚡ Standard | + +**Tip**: Beginners can start with text-only templates for instant zero-barrier experience! + +**Alternative**: If you need AI images but don't want local ComfyUI, use RunningHub cloud service. ### Q: Which LLMs are supported? diff --git a/docs/zh/faq.md b/docs/zh/faq.md index 1c47db2..137506d 100644 --- a/docs/zh/faq.md +++ b/docs/zh/faq.md @@ -22,7 +22,16 @@ curl -LsSf https://astral.sh/uv/install.sh | sh ### Q: 必须要配置 ComfyUI 吗？ -不一定。你可以使用 RunningHub 云端服务，无需本地部署。 +**不一定**，取决于您选择的模板类型： + +| 模板类型 | ComfyUI | 适用场景 | 生成速度 | +|---------|---------|---------|----------| +| 纯文本模板
（如 `simple.html`） | ❌ 不需要 | 文字金句、公告、阅读提示 | ⚡⚡⚡ 极快 | +| AI 配图模板
（如 `default.html`） | ✅ 需要 | 图文并茂的丰富内容 | ⚡ 标准 | + +**推荐**：新手可以从纯文本模板开始，零门槛快速体验！ + +**其他选项**：如果需要 AI 配图但不想本地部署 ComfyUI，可以使用 RunningHub 云端服务。 ### Q: 支持哪些 LLM？ diff --git a/pixelle_video/pipelines/custom.py b/pixelle_video/pipelines/custom.py index aa4d6fa..1724193 100644 --- a/pixelle_video/pipelines/custom.py +++ b/pixelle_video/pipelines/custom.py @@ -32,10 +32,24 @@ class CustomPipeline(BasePipeline): You can customize: - Content processing logic - Narration generation strategy - - Image prompt generation + - Image prompt generation (conditional based on template) - Frame composition - Video assembly + KEY OPTIMIZATION: Conditional Image Generation + ----------------------------------------------- + This pipeline supports automatic detection of template image requirements. + If your template doesn't use {{image}}, the entire image generation pipeline + can be skipped, providing: + ⚡ Faster generation (no image API calls) + 💰 Lower cost (no LLM calls for image prompts) + 🚀 Reduced dependencies (no ComfyUI needed for text-only videos) + + Usage patterns: + 1. Text-only videos: Use templates/1080x1920/simple.html + 2. AI-generated images: Use templates with {{image}} placeholder + 3. Custom logic: Modify template or override the detection logic in your subclass + Example usage: # 1. Create your own pipeline by copying this file # 2. Modify the __call__ method with your custom logic @@ -90,6 +104,11 @@ class CustomPipeline(BasePipeline): Returns: VideoGenerationResult + + Image Generation Logic: + - If template has {{image}} → automatically generates images + - If template has no {{image}} → skips image generation (faster, cheaper) + - To customize: Override the template_requires_image logic in your subclass """ logger.info("Starting CustomPipeline") logger.info(f"Input text length: {len(text)} chars") @@ -114,6 +133,22 @@ class CustomPipeline(BasePipeline): user_specified_output = output_path output_path = get_task_final_video_path(task_id) + # ========== Step 0.5: Check template requirements ========== + # Detect if template requires {{image}} parameter + # This allows skipping the entire image generation pipeline for text-only templates + from pixelle_video.services.frame_html import HTMLFrameGenerator + from pixelle_video.utils.template_util import resolve_template_path + + template_path = resolve_template_path(frame_template) + generator = HTMLFrameGenerator(template_path) + template_requires_image = generator.requires_image() + + if template_requires_image: + logger.info(f"📸 Template requires image generation") + else: + logger.info(f"⚡ Template does not require images - skipping image generation pipeline") + logger.info(f" 💡 Benefits: Faster generation + Lower cost + No ComfyUI dependency") + # ========== Step 1: Process content (CUSTOMIZE THIS) ========== self._report_progress(progress_callback, "processing_content", 0.10) @@ -138,29 +173,37 @@ class CustomPipeline(BasePipeline): logger.info(f"Generated {len(narrations)} narrations") - # ========== Step 2: Generate image prompts (CUSTOMIZE THIS) ========== + # ========== Step 2: Generate image prompts (CONDITIONAL - CUSTOMIZE THIS) ========== self._report_progress(progress_callback, "generating_image_prompts", 0.25) - # Example: Generate image prompts using LLM - from pixelle_video.utils.content_generators import generate_image_prompts - - image_prompts = await generate_image_prompts( - self.llm, - narrations=narrations, - min_words=30, - max_words=60 - ) - - # Example: Apply custom prompt prefix - from pixelle_video.utils.prompt_helper import build_image_prompt - custom_prefix = "cinematic style, professional lighting" # Customize this - - final_image_prompts = [] - for base_prompt in image_prompts: - final_prompt = build_image_prompt(base_prompt, custom_prefix) - final_image_prompts.append(final_prompt) - - logger.info(f"Generated {len(final_image_prompts)} image prompts") + # IMPORTANT: Check if template actually needs images + # If your template doesn't use {{image}}, you can skip this entire step! + if template_requires_image: + # Template requires images - generate image prompts using LLM + from pixelle_video.utils.content_generators import generate_image_prompts + + image_prompts = await generate_image_prompts( + self.llm, + narrations=narrations, + min_words=30, + max_words=60 + ) + + # Example: Apply custom prompt prefix + from pixelle_video.utils.prompt_helper import build_image_prompt + custom_prefix = "cinematic style, professional lighting" # Customize this + + final_image_prompts = [] + for base_prompt in image_prompts: + final_prompt = build_image_prompt(base_prompt, custom_prefix) + final_image_prompts.append(final_prompt) + + logger.info(f"✅ Generated {len(final_image_prompts)} image prompts") + else: + # Template doesn't need images - skip image generation entirely + final_image_prompts = [None] * len(narrations) + logger.info(f"⚡ Skipped image prompt generation (template doesn't need images)") + logger.info(f" 💡 Savings: {len(narrations)} LLM calls + {len(narrations)} image generations") # ========== Step 3: Create storyboard ========== config = StoryboardConfig( @@ -317,8 +360,8 @@ class CustomPipeline(BasePipeline): # ==================== Usage Examples ==================== """ -Example 1: Register and use custom pipeline ----------------------------------------- +Example 1: Text-only video (no AI image generation) +--------------------------------------------------- from pixelle_video import pixelle_video from pixelle_video.pipelines.custom import CustomPipeline @@ -328,15 +371,27 @@ await pixelle_video.initialize() # Register custom pipeline pixelle_video.pipelines["my_custom"] = CustomPipeline(pixelle_video) -# Use it +# Use text-only template - no image generation! result = await pixelle_video.generate_video( - text="Your input content here", + text="Your content here", pipeline="my_custom", - custom_param_example="custom_value" + frame_template="1080x1920/simple.html" # Template without {{image}} ) +# Benefits: ⚡ Fast, 💰 Cheap, 🚀 No ComfyUI needed -Example 2: Create your own pipeline class +Example 2: AI-generated image video +--------------------------------------------------- +# Use template with {{image}} - automatic image generation +result = await pixelle_video.generate_video( + text="Your content here", + pipeline="my_custom", + frame_template="1080x1920/default.html" # Template with {{image}} +) +# Will automatically generate images via LLM + ComfyUI + + +Example 3: Create your own pipeline class ---------------------------------------- from pixelle_video.pipelines.custom import CustomPipeline @@ -351,7 +406,7 @@ class MySpecialPipeline(CustomPipeline): return result -Example 3: Inline custom pipeline +Example 4: Inline custom pipeline ---------------------------------------- from pixelle_video.pipelines.base import BasePipeline diff --git a/pixelle_video/pipelines/standard.py b/pixelle_video/pipelines/standard.py index 3539194..0318dcc 100644 --- a/pixelle_video/pipelines/standard.py +++ b/pixelle_video/pipelines/standard.py @@ -250,6 +250,14 @@ class StandardPipeline(BasePipeline): created_at=datetime.now() ) + # ========== Step 0.8: Check template requirements ========== + template_requires_image = self._check_template_requires_image(config.frame_template) + if template_requires_image: + logger.info(f"📸 Template requires image generation") + else: + logger.info(f"⚡ Template does not require images - skipping image generation pipeline") + logger.info(f" 💡 Benefits: Faster generation + Lower cost + No ComfyUI dependency") + try: # ========== Step 1: Generate/Split narrations ========== if mode == "generate": @@ -268,54 +276,61 @@ class StandardPipeline(BasePipeline): logger.info(f"✅ Split script into {len(narrations)} segments (by lines)") logger.info(f" Note: n_scenes={n_scenes} is ignored in fixed mode") - # ========== Step 2: Generate image prompts ========== - self._report_progress(progress_callback, "generating_image_prompts", 0.15) - - # Override prompt_prefix if provided - original_prefix = None - if prompt_prefix is not None: - image_config = self.core.config.get("comfyui", {}).get("image", {}) - original_prefix = image_config.get("prompt_prefix") - image_config["prompt_prefix"] = prompt_prefix - logger.info(f"Using custom prompt_prefix: '{prompt_prefix}'") - - try: - # Create progress callback wrapper for image prompt generation - def image_prompt_progress(completed: int, total: int, message: str): - batch_progress = completed / total if total > 0 else 0 - overall_progress = 0.15 + (batch_progress * 0.15) - self._report_progress( - progress_callback, - "generating_image_prompts", - overall_progress, - extra_info=message + # ========== Step 2: Generate image prompts (conditional) ========== + if template_requires_image: + self._report_progress(progress_callback, "generating_image_prompts", 0.15) + + # Override prompt_prefix if provided + original_prefix = None + if prompt_prefix is not None: + image_config = self.core.config.get("comfyui", {}).get("image", {}) + original_prefix = image_config.get("prompt_prefix") + image_config["prompt_prefix"] = prompt_prefix + logger.info(f"Using custom prompt_prefix: '{prompt_prefix}'") + + try: + # Create progress callback wrapper for image prompt generation + def image_prompt_progress(completed: int, total: int, message: str): + batch_progress = completed / total if total > 0 else 0 + overall_progress = 0.15 + (batch_progress * 0.15) + self._report_progress( + progress_callback, + "generating_image_prompts", + overall_progress, + extra_info=message + ) + + # Generate base image prompts + base_image_prompts = await generate_image_prompts( + self.llm, + narrations=narrations, + min_words=min_image_prompt_words, + max_words=max_image_prompt_words, + progress_callback=image_prompt_progress ) + + # Apply prompt prefix + from pixelle_video.utils.prompt_helper import build_image_prompt + image_config = self.core.config.get("comfyui", {}).get("image", {}) + prompt_prefix_to_use = prompt_prefix if prompt_prefix is not None else image_config.get("prompt_prefix", "") + + image_prompts = [] + for base_prompt in base_image_prompts: + final_prompt = build_image_prompt(base_prompt, prompt_prefix_to_use) + image_prompts.append(final_prompt) + + finally: + # Restore original prompt_prefix + if original_prefix is not None: + image_config["prompt_prefix"] = original_prefix - # Generate base image prompts - base_image_prompts = await generate_image_prompts( - self.llm, - narrations=narrations, - min_words=min_image_prompt_words, - max_words=max_image_prompt_words, - progress_callback=image_prompt_progress - ) - - # Apply prompt prefix - from pixelle_video.utils.prompt_helper import build_image_prompt - image_config = self.core.config.get("comfyui", {}).get("image", {}) - prompt_prefix_to_use = prompt_prefix if prompt_prefix is not None else image_config.get("prompt_prefix", "") - - image_prompts = [] - for base_prompt in base_image_prompts: - final_prompt = build_image_prompt(base_prompt, prompt_prefix_to_use) - image_prompts.append(final_prompt) - - finally: - # Restore original prompt_prefix - if original_prefix is not None: - image_config["prompt_prefix"] = original_prefix - - logger.info(f"✅ Generated {len(image_prompts)} image prompts") + logger.info(f"✅ Generated {len(image_prompts)} image prompts") + else: + # Skip image prompt generation + image_prompts = [None] * len(narrations) + self._report_progress(progress_callback, "preparing_frames", 0.15) + logger.info(f"⚡ Skipped image prompt generation (template doesn't need images)") + logger.info(f" 💡 Savings: {len(narrations)} LLM calls + {len(narrations)} image generations") # ========== Step 3: Create frames ========== for i, (narration, image_prompt) in enumerate(zip(narrations, image_prompts)): @@ -418,4 +433,30 @@ class StandardPipeline(BasePipeline): except Exception as e: logger.error(f"❌ Video generation failed: {e}") raise + + def _check_template_requires_image(self, frame_template: str) -> bool: + """ + Check if template requires image generation + + This is checked at pipeline level to avoid unnecessary: + - LLM calls (generating image_prompts) + - Image generation API calls + - ComfyUI dependency + + Args: + frame_template: Template path (e.g., "1080x1920/default.html") + + Returns: + True if template contains {{image}}, False otherwise + """ + from pixelle_video.services.frame_html import HTMLFrameGenerator + from pixelle_video.utils.template_util import resolve_template_path + + template_path = resolve_template_path(frame_template) + generator = HTMLFrameGenerator(template_path) + + requires = generator.requires_image() + logger.debug(f"Template '{frame_template}' requires_image={requires}") + + return requires diff --git a/pixelle_video/prompts/README.md b/pixelle_video/prompts/README.md deleted file mode 100644 index 2e973a1..0000000 --- a/pixelle_video/prompts/README.md +++ /dev/null @@ -1,99 +0,0 @@ -# Prompts Directory - -Centralized prompt management for all LLM interactions in Pixelle-Video. - -## Structure - -Each prompt is in its own file for easy maintenance and modification: - -``` -prompts/ -├── __init__.py # Exports all builder functions -├── topic_narration.py # Generate narrations from topic -├── content_narration.py # Extract/refine narrations from content -├── script_split.py # Split fixed script into segments -├── title_generation.py # Generate video title from content -├── image_generation.py # Generate image prompts from narrations -└── style_conversion.py # Convert style description to image prompt -``` - -## Usage - -All builder functions are exported from the package root: - -```python -from pixelle_video.prompts import ( - build_topic_narration_prompt, - build_content_narration_prompt, - build_script_split_prompt, - build_title_generation_prompt, - build_image_prompt_prompt, - build_style_conversion_prompt, -) -``` - -## Prompt Files - -### Narration Prompts - -1. **topic_narration.py** - - Purpose: Generate engaging narrations from a topic/theme - - Input: topic, n_storyboard, min_words, max_words - - Output: JSON with narrations array - -2. **content_narration.py** - - Purpose: Extract and refine narrations from user content - - Input: content, n_storyboard, min_words, max_words - - Output: JSON with narrations array - -3. **script_split.py** - - Purpose: Split fixed script into natural segments (no modification) - - Input: script, min_words (reference), max_words (reference) - - Output: JSON with narrations array - -4. **title_generation.py** - - Purpose: Generate short, attractive video title - - Input: content, max_length - - Output: Plain text title - -### Image Prompts - -5. **image_generation.py** - - Purpose: Generate English image prompts from narrations - - Input: narrations, min_words, max_words, style_preset/style_description - - Output: JSON with image_prompts array - - Contains: IMAGE_STYLE_PRESETS dictionary - -6. **style_conversion.py** - - Purpose: Convert custom style description to English image prompt - - Input: description (any language) - - Output: Plain text English image prompt - -## Modifying Prompts - -To modify a prompt: - -1. Locate the relevant file (e.g., `topic_narration.py`) -2. Edit the prompt constant (e.g., `TOPIC_NARRATION_PROMPT`) -3. Changes take effect immediately (no need to modify service code) - -## Adding New Prompts - -To add a new prompt: - -1. Create a new file (e.g., `my_new_prompt.py`) -2. Define the prompt constant and builder function -3. Export the builder function in `__init__.py` -4. Use it in service code: - ```python - from pixelle_video.prompts import build_my_new_prompt - ``` - -## Design Principles - -- **One File, One Prompt**: Each prompt has its own file for clarity -- **Builder Functions**: Each file exports a `build_*_prompt()` function -- **Centralized Exports**: All builders are exported from `__init__.py` -- **Consistent Format**: All prompts follow similar structure and style -- **Easy Maintenance**: Modify prompts without touching service code - diff --git a/pixelle_video/services/frame_html.py b/pixelle_video/services/frame_html.py index 133332d..309a1aa 100644 --- a/pixelle_video/services/frame_html.py +++ b/pixelle_video/services/frame_html.py @@ -57,6 +57,22 @@ class HTMLFrameGenerator: self._check_linux_dependencies() logger.debug(f"Loaded HTML template: {template_path} (size: {self.width}x{self.height})") + def requires_image(self) -> bool: + """ + Detect if template requires {{image}} parameter + + This method checks if the template uses the {{image}} variable. + If the template doesn't use images, the entire image generation + pipeline can be skipped, significantly improving: + - Generation speed (no image generation API calls) + - Cost efficiency (no LLM calls for image prompts) + - Dependency requirements (no ComfyUI needed) + + Returns: + True if template contains {{image}}, False otherwise + """ + return '{{image}}' in self.template + def _check_linux_dependencies(self): """Check Linux system dependencies and warn if missing""" if os.name != 'posix': @@ -403,7 +419,7 @@ class HTMLFrameGenerator: # Replace variables in HTML (supports DSL syntax: {{param:type=default}}) html = self._replace_parameters(self.template, context) - logger.info(f"html--->{html}") + logger.debug(f"html--->{html}") # Use provided output path or auto-generate if output_path is None: # Fallback: auto-generate (for backward compatibility) diff --git a/pixelle_video/services/frame_processor.py b/pixelle_video/services/frame_processor.py index 74d6f13..1077b0c 100644 --- a/pixelle_video/services/frame_processor.py +++ b/pixelle_video/services/frame_processor.py @@ -56,6 +56,9 @@ class FrameProcessor: frame_num = frame.index + 1 + # Determine if this frame needs image generation + needs_image = frame.image_prompt is not None + try: # Step 1: Generate audio (TTS) if progress_callback: @@ -69,23 +72,27 @@ class FrameProcessor: )) await self._step_generate_audio(frame, config) - # Step 2: Generate image (ComfyKit) - if progress_callback: - progress_callback(ProgressEvent( - event_type="frame_step", - progress=0.25, - frame_current=frame_num, - frame_total=total_frames, - step=2, - action="image" - )) - await self._step_generate_image(frame, config) + # Step 2: Generate image (conditional) + if needs_image: + if progress_callback: + progress_callback(ProgressEvent( + event_type="frame_step", + progress=0.25, + frame_current=frame_num, + frame_total=total_frames, + step=2, + action="image" + )) + await self._step_generate_image(frame, config) + else: + frame.image_path = None + logger.debug(f" 2/4: Skipped image generation (not required by template)") # Step 3: Compose frame (add subtitle) if progress_callback: progress_callback(ProgressEvent( event_type="frame_step", - progress=0.50, + progress=0.50 if needs_image else 0.33, frame_current=frame_num, frame_total=total_frames, step=3, @@ -97,7 +104,7 @@ class FrameProcessor: if progress_callback: progress_callback(ProgressEvent( event_type="frame_step", - progress=0.75, + progress=0.75 if needs_image else 0.67, frame_current=frame_num, frame_total=total_frames, step=4, diff --git a/templates/1080x1920/simple.html b/templates/1080x1920/simple.html new file mode 100644 index 0000000..18a05ec --- /dev/null +++ b/templates/1080x1920/simple.html @@ -0,0 +1,278 @@ + + + + + + + + +
+ Background +
+ + +
+ + +
+
+
+
+
+ +
+ +
+
{{title}}
+
+
+ + +
+
+
"
+
{{text}}
+
"
+
+
+ + +
+
+
{{author=@Pixelle.AI}}
+
{{describe=Open Source Omnimodal AI Creative Agent}}
+
+
+ +
Text-Only Template
+
+
+
+ + + diff --git a/web/app.py b/web/app.py index eea3c20..79510c0 100644 --- a/web/app.py +++ b/web/app.py @@ -636,145 +636,6 @@ def main(): st.error(tr("tts.preview_failed", error=str(e))) logger.exception(e) - # ==================================================================== - # Image Generation Section - # ==================================================================== - with st.container(border=True): - st.markdown(f"**{tr('section.image')}**") - - # 1. ComfyUI Workflow selection - with st.expander(tr("help.feature_description"), expanded=False): - st.markdown(f"**{tr('help.what')}**") - st.markdown(tr("style.workflow_what")) - st.markdown(f"**{tr('help.how')}**") - st.markdown(tr("style.workflow_how")) - st.markdown(f"**{tr('help.note')}**") - st.markdown(tr("style.image_size_note")) - - # Get available workflows from pixelle_video (with source info) - workflows = pixelle_video.image.list_workflows() - - # Build options for selectbox - # Display: "image_flux.json - Runninghub" - # Value: "runninghub/image_flux.json" - workflow_options = [wf["display_name"] for wf in workflows] - workflow_keys = [wf["key"] for wf in workflows] - - # Default to first option (should be runninghub by sorting) - default_workflow_index = 0 - - # If user has a saved preference in config, try to match it - comfyui_config = config_manager.get_comfyui_config() - saved_workflow = comfyui_config["image"]["default_workflow"] - if saved_workflow and saved_workflow in workflow_keys: - default_workflow_index = workflow_keys.index(saved_workflow) - - workflow_display = st.selectbox( - "Workflow", - workflow_options if workflow_options else ["No workflows found"], - index=default_workflow_index, - label_visibility="collapsed", - key="image_workflow_select" - ) - - # Get the actual workflow key (e.g., "runninghub/image_flux.json") - if workflow_options: - workflow_selected_index = workflow_options.index(workflow_display) - workflow_key = workflow_keys[workflow_selected_index] - else: - workflow_key = "runninghub/image_flux.json" # fallback - - - # 2. Image size input - col1, col2 = st.columns(2) - with col1: - image_width = st.number_input( - tr('style.image_width'), - min_value=128, - value=1024, - step=1, - label_visibility="visible", - help=tr('style.image_width_help') - ) - with col2: - image_height = st.number_input( - tr('style.image_height'), - min_value=128, - value=1024, - step=1, - label_visibility="visible", - help=tr('style.image_height_help') - ) - - # 3. Prompt prefix input - # Get current prompt_prefix from config - current_prefix = comfyui_config["image"]["prompt_prefix"] - - # Prompt prefix input (temporary, not saved to config) - prompt_prefix = st.text_area( - tr('style.prompt_prefix'), - value=current_prefix, - placeholder=tr("style.prompt_prefix_placeholder"), - height=80, - label_visibility="visible", - help=tr("style.prompt_prefix_help") - ) - - # Style preview expander (similar to template preview) - with st.expander(tr("style.preview_title"), expanded=False): - # Test prompt input - test_prompt = st.text_input( - tr("style.test_prompt"), - value="a dog", - help=tr("style.test_prompt_help"), - key="style_test_prompt" - ) - - # Preview button - if st.button(tr("style.preview"), key="preview_style", use_container_width=True): - with st.spinner(tr("style.previewing")): - try: - from pixelle_video.utils.prompt_helper import build_image_prompt - - # Build final prompt with prefix - final_prompt = build_image_prompt(test_prompt, prompt_prefix) - - # Generate preview image (use user-specified size) - preview_image_path = run_async(pixelle_video.image( - prompt=final_prompt, - workflow=workflow_key, - width=int(image_width), - height=int(image_height) - )) - - # Display preview (support both URL and local path) - if preview_image_path: - st.success(tr("style.preview_success")) - - # Read and encode image - if preview_image_path.startswith('http'): - # URL - use directly - img_html = f'
Style Preview
' - else: - # Local file - encode as base64 - with open(preview_image_path, 'rb') as f: - img_data = base64.b64encode(f.read()).decode() - img_html = f'
Style Preview
' - - st.markdown(img_html, unsafe_allow_html=True) - - # Show the final prompt used - st.info(f"**{tr('style.final_prompt_label')}**\n{final_prompt}") - - # Show file path - st.caption(f"📁 {preview_image_path}") - else: - st.error(tr("style.preview_failed_general")) - except Exception as e: - st.error(tr("style.preview_failed", error=str(e))) - logger.exception(e) - - # ==================================================================== # Storyboard Template Section # ==================================================================== @@ -866,6 +727,11 @@ def main(): generator_for_params = HTMLFrameGenerator(template_path_for_params) custom_params_for_video = generator_for_params.parse_template_parameters() + # Detect if template requires image generation + template_requires_image = generator_for_params.requires_image() + # Store in session state for Image Section to read + st.session_state['template_requires_image'] = template_requires_image + custom_values_for_video = {} if custom_params_for_video: st.markdown("📝 " + tr("template.custom_parameters")) @@ -1004,7 +870,163 @@ def main(): except Exception as e: st.error(tr("template.preview_failed", error=str(e))) logger.exception(e) - + + # ==================================================================== + # Image Generation Section (conditional based on template) + # ==================================================================== + # Check if current template requires image generation + if st.session_state.get('template_requires_image', True): + # Template requires images - show full Image Section + with st.container(border=True): + st.markdown(f"**{tr('section.image')}**") + + # 1. ComfyUI Workflow selection + with st.expander(tr("help.feature_description"), expanded=False): + st.markdown(f"**{tr('help.what')}**") + st.markdown(tr("style.workflow_what")) + st.markdown(f"**{tr('help.how')}**") + st.markdown(tr("style.workflow_how")) + st.markdown(f"**{tr('help.note')}**") + st.markdown(tr("style.image_size_note")) + + # Get available workflows from pixelle_video (with source info) + workflows = pixelle_video.image.list_workflows() + + # Build options for selectbox + # Display: "image_flux.json - Runninghub" + # Value: "runninghub/image_flux.json" + workflow_options = [wf["display_name"] for wf in workflows] + workflow_keys = [wf["key"] for wf in workflows] + + # Default to first option (should be runninghub by sorting) + default_workflow_index = 0 + + # If user has a saved preference in config, try to match it + comfyui_config = config_manager.get_comfyui_config() + saved_workflow = comfyui_config["image"]["default_workflow"] + if saved_workflow and saved_workflow in workflow_keys: + default_workflow_index = workflow_keys.index(saved_workflow) + + workflow_display = st.selectbox( + "Workflow", + workflow_options if workflow_options else ["No workflows found"], + index=default_workflow_index, + label_visibility="collapsed", + key="image_workflow_select" + ) + + # Get the actual workflow key (e.g., "runninghub/image_flux.json") + if workflow_options: + workflow_selected_index = workflow_options.index(workflow_display) + workflow_key = workflow_keys[workflow_selected_index] + else: + workflow_key = "runninghub/image_flux.json" # fallback + + + # 2. Image size input + col1, col2 = st.columns(2) + with col1: + image_width = st.number_input( + tr('style.image_width'), + min_value=128, + value=1024, + step=1, + label_visibility="visible", + help=tr('style.image_width_help') + ) + with col2: + image_height = st.number_input( + tr('style.image_height'), + min_value=128, + value=1024, + step=1, + label_visibility="visible", + help=tr('style.image_height_help') + ) + + # 3. Prompt prefix input + # Get current prompt_prefix from config + current_prefix = comfyui_config["image"]["prompt_prefix"] + + # Prompt prefix input (temporary, not saved to config) + prompt_prefix = st.text_area( + tr('style.prompt_prefix'), + value=current_prefix, + placeholder=tr("style.prompt_prefix_placeholder"), + height=80, + label_visibility="visible", + help=tr("style.prompt_prefix_help") + ) + + # Style preview expander (similar to template preview) + with st.expander(tr("style.preview_title"), expanded=False): + # Test prompt input + test_prompt = st.text_input( + tr("style.test_prompt"), + value="a dog", + help=tr("style.test_prompt_help"), + key="style_test_prompt" + ) + + # Preview button + if st.button(tr("style.preview"), key="preview_style", use_container_width=True): + with st.spinner(tr("style.previewing")): + try: + from pixelle_video.utils.prompt_helper import build_image_prompt + + # Build final prompt with prefix + final_prompt = build_image_prompt(test_prompt, prompt_prefix) + + # Generate preview image (use user-specified size) + preview_image_path = run_async(pixelle_video.image( + prompt=final_prompt, + workflow=workflow_key, + width=int(image_width), + height=int(image_height) + )) + + # Display preview (support both URL and local path) + if preview_image_path: + st.success(tr("style.preview_success")) + + # Read and encode image + if preview_image_path.startswith('http'): + # URL - use directly + img_html = f'
Style Preview
' + else: + # Local file - encode as base64 + with open(preview_image_path, 'rb') as f: + img_data = base64.b64encode(f.read()).decode() + img_html = f'
Style Preview
' + + st.markdown(img_html, unsafe_allow_html=True) + + # Show the final prompt used + st.info(f"**{tr('style.final_prompt_label')}**\n{final_prompt}") + + # Show file path + st.caption(f"📁 {preview_image_path}") + else: + st.error(tr("style.preview_failed_general")) + except Exception as e: + st.error(tr("style.preview_failed", error=str(e))) + logger.exception(e) + + + else: + # Template doesn't need images - show simplified message + with st.container(border=True): + st.markdown(f"**{tr('section.image')}**") + st.info("ℹ️ " + tr("image.not_required")) + st.caption(tr("image.not_required_hint")) + + # Set default values for later use + workflow_key = None + image_width = 1024 + image_height = 1024 + prompt_prefix = "" + + # ======================================================================== # Right Column: Generate Button + Progress + Video Preview # ======================================================================== diff --git a/web/i18n/locales/en_US.json b/web/i18n/locales/en_US.json index b7c2b06..ab0df9d 100644 --- a/web/i18n/locales/en_US.json +++ b/web/i18n/locales/en_US.json @@ -102,6 +102,9 @@ "template.preview_caption": "Template Preview: {template}", "template.custom_parameters": "Custom Parameters", + "image.not_required": "Current template does not require image generation", + "image.not_required_hint": "The selected template is text-only and does not need images. Benefits: ⚡ Faster generation 💰 Lower cost", + "video.title": "🎬 Video Settings", "video.frames": "Scenes", "video.frames_help": "More scenes = longer video", diff --git a/web/i18n/locales/zh_CN.json b/web/i18n/locales/zh_CN.json index 187c377..9d3e801 100644 --- a/web/i18n/locales/zh_CN.json +++ b/web/i18n/locales/zh_CN.json @@ -102,6 +102,9 @@ "template.preview_caption": "模板预览：{template}", "template.custom_parameters": "自定义参数", + "image.not_required": "当前模板不需要插图生成", + "image.not_required_hint": "您选择的模板是纯文本模板，无需生成图片。这将：⚡ 加快生成速度 💰 降低生成成本", + "video.title": "🎬 视频设置", "video.frames": "分镜数", "video.frames_help": "更多分镜 = 更长视频",